Expense constrained bidder optimization in repeated auctions Ramki Gummadi Stanford University (Based on joint work with P. Key and A. Proutiere)
Overview Introduction/Motivation Budgeted Second Price Auctions A General Online Budgeting Framework Optimal Bids for Micro-Value Auctions Conclusion
Three Aspects of Sponsored Search 1.Sequential setting. 2. Micro-transactions per auction. 3. The long tail of advertisers is expense constrained.
Motivation: Expense Constraints Payments are explicit, but valuations are abstract. Significantly alters bidding behavior. Critical for advertisers in the long tail.
Modeling Expense Constraints Balance time T0 B
Modeling Expense Constraints Stochastic fluctuations could cause spend rate different from target. Balance time T0 B
Modeling Expense Constraints “…the nature of what this budget limit means for the bidders themselves is somewhat of a mystery. There seems to be some risk control element to it, some purely administrative element to it, some bounded- rationality element to it, and more…” -- “Theory research at google”, SIGACT News, 2008.
Modeling Expense Constraints Balance time 0 B
Responsibility for expense constraints Auctioneer Bidder Bids fixed -- Auction entry throttled. Bids adjusted dynamically. Online bipartite matching between queries and bidders. Online knapsack type problems. Expense constraints = fixed budget. Possible to model more general expense constraints.
Bid optimization
Modeling aspects Expense constraints include a running balance constraint together with a fixed income per time slot. Random i.i.d. environment models aggregate statistics. -- observable and non-observable components. Bids are lower because any money saved can instead be used to buy a cheaper auction in the future. Objective function is infinite horizon expected utility, but with a discount factor that models limited patience.
Preview
Preview: Optimal Shading factors
Overview Introduction Budgeted Second Price auctions A General Online Budgeting Framework Optimal Bids for Micro-Value Auctions Conclusion
Model: Budgeted Second Price
The Value Function
But boundary conditions can not be inferred from the DP argument. Current auction Loss Win
Future opportunity cost Characterization of value function
Value Iteration:
Limiting case: micro-value auctions
Overview Introduction Budgeted Second Price Auctions A General Online Budgeting Framework Optimal Bids for Micro-Value Auctions Conclusion
General Online Budgeting Model Decision Maker Unobservable
Ex1: Second Price Auction
Ex2: GSP Auction Click events for L slots
Overview Introduction Budgeted Second Price Auctions A General Online Budgeting Framework Optimal Bids for Micro-Value Auctions Conclusion
Notation:
Theorem
Application to Second Price Auctions
Second Price Auction Example Opponents bid p Value functions
Optimal bid i.e., Static SP with shaded valuation:
Optimal Scaling factor
Optimal Bid: GSP Static GSP with “virtual valuation”:
Proof Overview Next 2 slides
time B(t) B Play U*
Overview Introduction MDP for budgeted SP auctions A General Online Budgeting Framework Optimal Bids for Micro-Value Auctions Conclusion
Stationarity in large markets
Conclusion A two parameter model for expense constraints in online budgeting problems. Optimal bid can be mapped to static auction with a shaded virtual valuation. Paper has more contents: MFE analysis and a finite horizon model.